Design, Elaboration, and Characterization of an Immunosensor for the Detection of a Fungal Toxin in Foodstuff Analyses

This work describes the complete elaboration of an immunosensor for the detection of the fungal B1 aflatoxin (AFB1). In a first step, a system made of three screen-printed electrodes (SPEs) was manufactured using gold, silver/silver chloride, and carbon pastes. Raman spectroscopy showed that the thermal treatment applied to the electrodes enabled a strong decrease in the amount of undesirable organic molecules for each paste. Atomic Force Microscopy was also used to reveal the morphology of the electrode surfaces. In a second step, an autonomous and cheap electronic system was designed for the control of the sensor and electrochemical measurements, showing current variations significantly higher than those observed with a commercial system. In a last step, the gold working electrode of this system was functionalized by a simple self-assembly method, optimized in a previous work, with a molecular architecture including an antibody recognizing specifically AFB1. The complete device was finally realized by combining the SPEs and the electronic platform. The resulting setup was able to detect AFB1 toxin in a buffer with an LOD of about 50 fg/mL. It was then applied to the detection of AFB1 in rice milk, a more realistic medium comparable with those met in an agrifood context. The electrochemical detection of AFB1 was possible in a range of concentration between 0.5 pg/mL and 2.5 pg/mL, with the sensor behaving linearly in this range.

Traditional electrochemical cells are bulky and expensive and require real known-how to be handled. Moreover, some of them include films or drop electrodes made of mercury, a highly toxic metal, which necessitates laborious cleaning processes. The limitations imply both to consider other materials and to miniaturize the whole system to obtain an easy-touse microsensor [9,10]. This miniaturized laboratory approach combined to "in-the-field" experiments [9] also reduces sample volumes and costs [9,10] and makes multidetection possible [11]. Miniaturized set-ups include lab-on-chip and point-of-care devices [7,12], which are successfully applied to the detection of multiple biomarkers. They are often based on measurements using single strip three-electrode systems, ultramicroelectrodes, electrode arrays, ring and paper-based electrodes [13,14], printed batch injection analysis

Materials
A screen printer from AUREL SPA (Modigliana, Italy) was used for screen printing. Before their functionalization, electrodes were stored protected from dust and kept at a constant temperature of 20 • C. Functionalized ones were kept in a TBS solution including KCl (2 mM) and used rapidly. NI-USB 6009 data acquisition card (DAQ) and TL081 and TL084 operational amplifiers (AO) used for the conception of our homemade electronic system were purchased from National Instruments France (Colombes, France) and RS Components SAS (Beauvais, France), respectively. Two commercial electrochemical sensor platforms (Mini PSAT 910 from MetrOhm, Villebon-sur-Yvette, France; reference KS510 from Kanichi, Warrington, UK) were used for comparison.

Methods
AFM measurements were carried out with a Bioscope II setup (Veeco-Bruker, Santa Barbara, CA, USA) equipped with an IX71 Olympus inverted optical microscope coupled to a Nanoscope V controller. Images were obtained in tapping mode, using silicon cantilevers with a spring constant around 40 N/m and NCLV tips from Bruker, at a resolution of 512 × 512 pixels and with a scan rate between 0.5 and 1.0 Hz. Data presented below are typical from measurements obtained with 12 samples (6 samples before cleaning and 6 samples after treatment). It is noteworthy that before fritting, imaging was impossible since the pastes were viscous with a thick texture.
Raman spectra were obtained with a DXR setup from Thermo Scientific (Illkirch-Graffenstaden, France), running with the Omnic 8.2 software. Spectra were recorded between 3500 and 100 cm −1 using a laser diode emitting at a wavelength of 790 nm, focused on the sample by a ×50 objective. Raman spectra shown below were typically obtained from the addition of 10 spectra (a few seconds each).
Electrochemical measurements were achieved using the PalmSens 4 potentiostat from PalmSens (Houten, The Netherlands). Cyclic voltammetry scans were obtained in the presence of the K 4 [Fe(CN) 6 ] 3− /K 3 [Fe(CN) 6 ] 4− redox couple (10 mM in TBS), used as electrochemical probe for signal amplification, at a scan rate of 100 mV/s. The potential interval was adapted to the redox couple, from −0.4 V to +0.6 V. Measurements were performed using our sensors and screen-printed sensors purchased from MetrOhm (Villebon-sur-Yvette, France). Electrodes were similar to ours for an easy comparison.

Design and Printing Process of Electrodes
The first step of SPE design was to define the most convenient geometry for further amperometric measurements. We used the Inkscape software (version 0.92.1) using scalable vector graphics (SVG) image coding.
As shown in Figure 1a, the working electrode has the geometry of a microdisk, similar to a previous system developed in the same context [52], characterized by a radius r (0 ≤ r ≤ R 1 ) and a thickness z. Thus, it can diffuse current only in both (Or) and (Oz) directions, depending on the ohmic drop of the whole sensor and the response time of the electron transfer between all the electrodes [53,54]. The diffusion equation with time t of this geometry is as follows: where C is the concentration (mol·cm −3 ) of the electroactive species in contact with the electrode, r (cm) is the distance from the center O of the disk, and D is the diffusion coefficient (cm 2 ·s −1 ) of the electroactive species.

Design and Printing Process of Electrodes and Chemical and Morphological Analysis
3.1.1. Design and Printing Process of Electrodes The first step of SPE design was to define the most convenient geometry for further amperometric measurements. We used the Inkscape software (version 0.92.1) using scalable vector graphics (SVG) image coding.
As shown in Figure 1a, the working electrode has the geometry of a microdisk, similar to a previous system developed in the same context [52], characterized by a radius r (0 ) and a thickness z. Thus, it can diffuse current only in both (Or) and (Oz) directions, depending on the ohmic drop of the whole sensor and the response time of the electron transfer between all the electrodes [53,54]. The diffusion equation with time t of this geometry is as follows: where C is the concentration (mol•cm −3 ) of the electroactive species in contact with the electrode, r (cm) is the distance from the center O of the disk, and D is the diffusion coefficient (cm 2 •s −1 ) of the electroactive species. The shape of the reference and counter electrodes ( Figure 1b) is a quarter of ring, with R1 and R2 as internal and external radii, respectively. This architecture makes it possible to quickly reach the electrochemical stabilization state. In such conditions, the current can be described by the following equation: where n is the mol number of electrons exchanged in the electrochemical reaction, and F is the Faraday constant (96,485 C•mol −1 ). The shape of the reference and counter electrodes ( Figure 1b) is a quarter of ring, with R 1 and R 2 as internal and external radii, respectively. This architecture makes it possible to quickly reach the electrochemical stabilization state. In such conditions, the current can be described by the following equation: where n is the mol number of electrons exchanged in the electrochemical reaction, and F is the Faraday constant (96,485 C·mol −1 ). Concerning the counter electrode, its surface must be larger than the surface of the reference one in order to minimize the current density at the level of this electrode. Therefore, we chose to double the surface of the counter electrode as compared to the reference one.
Microelectrodes were also integrated on the same chip to minimize the hysteresis (signal/noise) in the electrochemical system. This consequently made it possible to increase the sensitivity of the sensor and to lower its detection limit. Such systems have also other advantages over macro and conventional systems, such as an increased mass transport, a faster response time, and a decreased resistance of the solution. In addition, multiintegrated microelectrodes have the ability to increase the current transfer over individual ones while respecting the mass diffusion and the ohmic drop [55][56][57]. Based on these characteristics, we modeled our matrix of microsensors combining two active surfaces, which share the same counter and reference electrodes (Figure 1c).
Regarding the chemical composition of the electrode, gold was chosen for the working one since it keeps a high capacity to conduct current, even when self-assembled layers are added [18,20]. For the reference electrode, we chose a Ag/AgCl paste rather than applying a chlorination process on a silver ink [7,28], which could be responsible for a lack of stability or reproducibility (data not shown). Finally, we chose a carbon ink for the counter electrode as it is inert, not corrosive, cheap, and ecofriendly compared to the mercury bulky electrodes [19,20]. However, instead of applying carbon directly on the substrate's surface, we first deposited a layer of silver. This layer promotes the charge transfer and ensures the connection with the electronic system, since carbon has a strong ohmic resistivity. Figure 1c shows the scheme of the whole microelectrochemical chip.
Finally, we used a fabrication process detailed elsewhere [9,29,38] and briefly described here, consisting of the successive deposition of pastes. The first step consisted of printing the electrode's masks and then insulating them on stainless steel screens using negative photo emulsion sheets. The second step was the printing process itself, based on the deposition and spreading of pastes. This process was achieved using a squeegee moving across the screen stencil under an adapted pressure in order to apply pastes homogenously without breaking the metallic mesh. Between two consecutive deposition steps, a thermal treatment was applied to the last deposited paste. This treatment improved spreading and adhesion, prevented the paste from overflowing and finally solidified it. This fritting step is also essential to evaporate unwanted polymeric binders and other additives. The applied temperatures were as follows:

•
The first paste (gold one) required the most elevated temperature, above 800 • C, for 30 min, followed by a slower cooling for 150 min until ambient temperature; • The second paste (Ag/AgCl one) was heated at 150 • C for 15 min, the thermal treatment being followed by a slow cooling for 120 min; • The third and last one (carbon paste) was heated at 200 • C for 120 min, also followed by a cooling phase until ambient temperature.
The final manufactured product is presented in Figure 1c. Interferometric measurements [57] gave a typical thickness around 17 µm for carbon and gold electrodes and 10 µm for the Ag/AgCl one, indicating an efficient deposition protocol and homogenous layers.

Chemical and Morphological Analysis
Pastes contain polymer binders and solvents that could interfere with the chemicals used to functionalize electrodes or those involved in the detection itself. Their composition (given by the manufacturer) is described in Table 1. Table 1. Chemical composition of pastes used for electrodes (data from manufacturers).

Paste Chemical Composition
Gold Gold: 80-90%; Terpineol: 10-20%; Glass frit containing 1-10% of lead; Cadmium oxide: 0. Therefore, electrodes must be free of such interfering compounds. Their elimination after heating can be assessed by comparing Raman spectra of pastes before and after the thermal treatment ( Figure 2):

•
Gold paste ( Figure 2a): Before heating, the Raman spectrum reveals the presence of different bands at 2960 cm −1 and in the 1500-900 cm −1 range, which could be assigned, at least partly, to terpineol, an aromatic molecule present in the paste. On the whole, these bands disappear after heating, being replaced by a broad band centered at 1250-1000 cm −1 and another one at 615 cm −1 that could be due to unidentified products coming from terpineol degradation [58]; • Ag/AgCl paste ( Figure 2b): Before heating, bands in a spectral range compatible with the presence of organic molecules are observed. In particular, the paste contains butyl di-glycol acetate rich in methylene groups that could be responsible for bands at 2930 cm −1 , 1450 cm −1 , and 1304 cm −1 . Skeletal C-C vibrations could give the band at 1120 cm −1 . After heating, bands are still observed but clearly reduced in intensity, suggesting a thermally induced degradation [59]; • Carbon paste ( Figure 2c): Before heating, the Raman spectrum shows the presence of two main bands at 1600 cm −1 and 1300 cm −1 , assigned to the presence of carbon [60], still observed (but with a lower resolution) after heating. The disappearance of other bands previously observed in the 900-500 cm −1 spectral range is probably a sign of degradation of impurities [61].  A complementary morphological characterization of the electrode's surface after heating was carried out using AFM. Figure 2 shows typical images differing according to the nature of electrodes.

•
Gold electrode ( Figure 2a): The surface is characterized by the presence of aggregates imperfectly merged and a rather large roughness reaching the µm range; • Ag/AgCl electrode ( Figure 2b): The surface is more homogenous than in the previous case, with a reduced roughness estimated at 400 nm in larger images. The zoomed images reveal the presence of small and merged aggregates homogenously distributed; • Carbon electrode (Figure 2c): Here also images show a grainy structure covering homogenously the surface, similar to the structures observed with glassy carbon substrates [62].

Design and Assembly of the Homemade Potentiostat
Cyclic voltammetry (CV) and chronoamperometry (CA) were then implemented in our biosensor. In both techniques, a potential is applied between the reference and working electrodes. The resulting current describes qualitatively and quantitatively the electronic behavior of the microchip and the electron transfer due to the biochemical reaction at the surface of the working electrode. The reference electrode must be protected from external noises or chemical reactions, which could modify the surface properties. In addition, a stable potential must be maintained, while the current value must be null. Consequently, a regulation system was added in order to maintain the suitable current distribution all over the system. The overall principle of the electrochemical platform is presented in Figure 3.

Design and Assembly of the Homemade Potentiostat
Cyclic voltammetry (CV) and chronoamperometry (CA) were then implemented in our biosensor. In both techniques, a potential is applied between the reference and working electrodes. The resulting current describes qualitatively and quantitatively the electronic behavior of the microchip and the electron transfer due to the biochemical reaction at the surface of the working electrode. The reference electrode must be protected from external noises or chemical reactions, which could modify the surface properties. In addition, a stable potential must be maintained, while the current value must be null. Consequently, a regulation system was added in order to maintain the suitable current distribution all over the system. The overall principle of the electrochemical platform is presented in Figure 3. Like other electronic systems, our homemade potentiostat also includes hardware and software parts:  The hardware part includes two blocks. Firstly, a NI-USB 6009 platform is used specifically as an interface between the sensor and the computing interface. Secondly, a conditioning card, designed in collaboration with the Electronic Department of LOMA, is used to regulate the signal of the reference electrode and the Input/Output (I/O) signals. Details concerning the card are provided below;  The software controls the generation and the acquisition of the I/O signals. It is developed using the graphical language LabVIEW, also used for the signal treatment, by integrating and applying numerical filters and amplifiers. The diagrams in Figure  4 illustrate the different steps followed for the potential application in CV and CA measurements. Simplified and numerical equations of Nernst are integrated in Lab-VIEW for this purpose.  Like other electronic systems, our homemade potentiostat also includes hardware and software parts:

•
The hardware part includes two blocks. Firstly, a NI-USB 6009 platform is used specifically as an interface between the sensor and the computing interface. Secondly, a conditioning card, designed in collaboration with the Electronic Department of LOMA, is used to regulate the signal of the reference electrode and the Input/Output (I/O) signals. Details concerning the card are provided below; • The software controls the generation and the acquisition of the I/O signals. It is developed using the graphical language LabVIEW, also used for the signal treatment, by integrating and applying numerical filters and amplifiers. The diagrams in Figure 4 illustrate the different steps followed for the potential application in CV and CA measurements. Simplified and numerical equations of Nernst are integrated in LabVIEW for this purpose.
The process is described below in the case of CV measurements. To ensure the generation of a potential alternating between given negative and positive values, we used the differential mode where the output potential E applied on the electrodes is defined as follows: In this equation, A O0 and A O1 are the analog outputs of the NI-USB 6009 DAQ. The software controls the generation and the acquisition of the I/O signals. It is developed using the graphical language LabVIEW, also used for the signal treatment, by integrating and applying numerical filters and amplifiers. The diagrams in Figure  4 illustrate the different steps followed for the potential application in CV and CA measurements. Simplified and numerical equations of Nernst are integrated in Lab-VIEW for this purpose.  The CV program contains a "for loop" block, necessary to control the workflow statement and to execute the number of cycles required for measurements. Three "while loop" blocks are incorporated in the main loop in order to define the functions applied to the sensor:

•
The first block is used to generate a triangular signal between two extreme potential values E beg and E end corresponding to the DAQ outputs A O0 and A O1 , respectively. Potential varying linearly and with a triangular shape between these two values it comes: where t is the time. The first equation corresponds to the potential increase from E beg to E end , whereas the second one corresponds to the potential decrease from E end and E beg . In these equations, k is the absolute value of the slope of each linear part of the triangular signal. It is thus possible to determine the time t 1 corresponding to one half of this signal: • The second block corresponds to the response time or the delay necessary to leave the biochemical process taking place on the active surface; • The third block uses the "DAQ assistant" to acquire the specific signals (current and voltage) to help save data and displaying the required CV waveform. We integrated a "Savitsky-Golay" filter modulus in order to numerically smooth the signal and to disable the noise-to-signal ratio without losing the initial data.
The conditioning card previously mentioned is described in Figure 5. The potential being generated in a differential mode, an AO (TL084) amplifier, referred to the ground, is set to measure this difference. Once the required potential is obtained, a common follower AO (TL081), connected to the reference electrode, maintains a negligible current during measurements in order to keep null the current passing through the reference electrode. The first stage is also connected to a second inverter amplifier, so that the measured potential is calculated as follows: where V ground is the mass potential, and V ref is the potential passing through the reference electrode.
Concerning the amplifier in association with the counter electrode, its fundamental role is to impose a differential fixed and adjustable potential between the working and the counter electrodes. This helps to maintain an electrical equilibrium all over the sensor and to protect it against an over potential. The last amplifier, a current-to-voltage con-verter, is connected to the working electrode (NI-USB 6009 DAQ being unable to acquire current data). Concerning the amplifier in association with the counter electrode, its fundamental role is to impose a differential fixed and adjustable potential between the working and the counter electrodes. This helps to maintain an electrical equilibrium all over the sensor and to protect it against an over potential. The last amplifier, a current-to-voltage converter, is connected to the working electrode (NI-USB 6009 DAQ being unable to acquire current data).
Validation of the electronic system was finally carried out using the dummy cell circuitry commercialized with the PalmSens potentiostat. As shown in Figure 6, the application of a linear potential between −0.4 V and +0.6 V reveals a linear and reversible current behavior. The application of Ohm's law makes it possible to obtain a resistance of 10 kΩ, corresponding to the value of the manufacturer and confirming the calibration of our system.  Validation of the electronic system was finally carried out using the dummy cell circuitry commercialized with the PalmSens potentiostat. As shown in Figure 6, the application of a linear potential between −0.4 V and +0.6 V reveals a linear and reversible current behavior. The application of Ohm's law makes it possible to obtain a resistance of 10 kΩ, corresponding to the value of the manufacturer and confirming the calibration of our system. Concerning the amplifier in association with the counter electrode, its fundamental role is to impose a differential fixed and adjustable potential between the working and the counter electrodes. This helps to maintain an electrical equilibrium all over the sensor and to protect it against an over potential. The last amplifier, a current-to-voltage converter, is connected to the working electrode (NI-USB 6009 DAQ being unable to acquire current data).
Validation of the electronic system was finally carried out using the dummy cell circuitry commercialized with the PalmSens potentiostat. As shown in Figure 6, the application of a linear potential between −0.4 V and +0.6 V reveals a linear and reversible current behavior. The application of Ohm's law makes it possible to obtain a resistance of 10 kΩ, corresponding to the value of the manufacturer and confirming the calibration of our system.

Electrochemical Interrogation of the Platform Performance
The first validation of our homemade electrochemical platform was carried out by comparing its voltammetric response to the response of a commercial SPE system. As shown in Figure 7a, both responses have the typical "duck" shape with two extrema characterized each by potential and current values (E pa , I pa ) and (E pc , I pc ) for oxidation and reduction, respectively. In addition, the electrochemical response of the homemade sensor stabilizes rapidly when applying potential current cycles (Figure 7b).

Electrochemical Interrogation of the Platform Performance
The first validation of our homemade electrochemical platform was carried out by comparing its voltammetric response to the response of a commercial SPE system. As shown in Figure 7a, both responses have the typical "duck" shape with two extrema characterized each by potential and current values (Epa, Ipa) and (Epc, Ipc) for oxidation and reduction, respectively. In addition, the electrochemical response of the homemade sensor stabilizes rapidly when applying potential current cycles (Figure 7b). The peak-to-peak separations ΔE = Epa − Epc are similar in both cases, estimated to 104 ± 6 mV and 102 ± 16 mV in the case of the commercial system and the homemade sensor, respectively, (average of three independent measurements). Both values are clearly above the theoretical value of 57 mV at 25 °C expected in the case of chemically and electrochemically stable reactions. The larger value of ΔE suggests that, in both cases, electron transfer reactions are slower than those theoretically expected, due to a higher barrier of charge transfer [7,53].
If ΔE values are similar, the amplitude of the current variation ΔI = Ipa − Ipc is, however, significantly higher in the case of the homemade sensor. In order to better understand this higher amplitude, we determined the electroactive areas AR − S of the working electrodes from CV diagrams shown in Figure 7, using the Randles-Sevčik equation: where Ip is the extremal current, n is the number of electrons involved in the redox reaction, D is the coefficient of diffusion of electroactive species, C is the concentration of the redox probe, and ν is the scan rate. It is then possible to determine electroactive areas AR − S from voltammetric responses and to compare them to the geometric areas Ageo of the electrodes determined from their dimensions [63,64]. The higher AR − S/Ageo ratio is displayed by our homemade microsensor: the electroactive area corresponds to 87% of the geometric surface, whereas it is in the range of 8-70% for commercial systems. This is likely due to the high ΔI values obtained with our sensor. Another crucial point is the scan rate. Figure 8 shows CV diagrams obtained with the homemade and the commercial sensors at scan rates ranging from 20 to 100 mV/s. For both sensors, Ipa and Ipc currents increase linearly with the square root of the scan rate, in agreement with the Randles-Sevčik equation (Equation 7), showing that redox species are diffusing freely. However, our homemade sensor is characterized by a higher slope (~18 μA/(V/s) 1/2 ) than the commercial one (~3 μA/(V/s) 1/2 ) (Figure 8c). Faster scan rates decrease the size of the diffusion layer (area close to the electrode where concentrations of species are different from their bulk concentration [63,65]), the electrochemical system becomes The peak-to-peak separations ∆E = E pa − E pc are similar in both cases, estimated to 104 ± 6 mV and 102 ± 16 mV in the case of the commercial system and the homemade sensor, respectively, (average of three independent measurements). Both values are clearly above the theoretical value of 57 mV at 25 • C expected in the case of chemically and electrochemically stable reactions. The larger value of ∆E suggests that, in both cases, electron transfer reactions are slower than those theoretically expected, due to a higher barrier of charge transfer [7,53].
If ∆E values are similar, the amplitude of the current variation ∆I = I pa − I pc is, however, significantly higher in the case of the homemade sensor. In order to better understand this higher amplitude, we determined the electroactive areas A R − S of the working electrodes from CV diagrams shown in Figure 7, using the Randles-Sevčik equation: I p = 2.69 × 10 5 n 3/2 A R−S CD 1/2 υ 1/2 (7) where I p is the extremal current, n is the number of electrons involved in the redox reaction, D is the coefficient of diffusion of electroactive species, C is the concentration of the redox probe, and ν is the scan rate. It is then possible to determine electroactive areas A R−S from voltammetric responses and to compare them to the geometric areas A geo of the electrodes determined from their dimensions [63,64]. The higher A R−S /A geo ratio is displayed by our homemade microsensor: the electroactive area corresponds to 87% of the geometric surface, whereas it is in the range of 8-70% for commercial systems. This is likely due to the high ∆I values obtained with our sensor. Another crucial point is the scan rate. Figure 8 shows CV diagrams obtained with the homemade and the commercial sensors at scan rates ranging from 20 to 100 mV/s. For both sensors, I pa and I pc currents increase linearly with the square root of the scan rate, in agreement with the Randles-Sevčik equation (Equation 7), showing that redox species are diffusing freely. However, our homemade sensor is characterized by a higher slope (~18 µA/(V/s) 1/2 ) than the commercial one (~3 µA/(V/s) 1/2 ) (Figure 8c). Faster scan rates decrease the size of the diffusion layer (area close to the electrode where concentrations of species are different from their bulk concentration [63,65]), the electrochemical system becomes more reversible, and the current increases [65]. This suggests that the diffusion layer is thinner in our homemade sensor. Varying the scan rate has also an effect on the ∆E peak-to-peak separation (Figure 8d). In both cases, a decrease in the ∆E value is on the whole observed at increasing square root of the scan rate. However, the homemade sensor reaches, at a scan rate of 20 mV/s, a value of 58 mV, thus approaching the theoretical value of 57 mV mentioned above, contrary to the commercial one. Consequently, our sensor reaches almost the perfect electrochemical reversibility at 20 mV/s. However, it is not possible to use this value during measurements since it does not make it possible to rapidly obtain a stable response from the sensor (data not shown). more reversible, and the current increases [65]. This suggests that the diffusion layer is thinner in our homemade sensor. Varying the scan rate has also an effect on the ΔE peakto-peak separation (Figure 8d). In both cases, a decrease in the ΔE value is on the whole observed at increasing square root of the scan rate. However, the homemade sensor reaches, at a scan rate of 20 mV/s, a value of 58 mV, thus approaching the theoretical value of 57 mV mentioned above, contrary to the commercial one. Consequently, our sensor reaches almost the perfect electrochemical reversibility at 20 mV/s. However, it is not possible to use this value during measurements since it does not make it possible to rapidly obtain a stable response from the sensor (data not shown).

Application of the Homemade Sensor to the Detection of the AFB1 Fungal Toxin
We first applied to the homemade sensor different concentrations of the Fe 3+ /Fe 2+ redox couple in the absence of functionalization on the working electrode. Three concentrations ranging from 1 mM to 10 mM were studied at a scan rate of 100 mV/s: the amplitude of the CV diagrams decreases proportionally to the concentration of the probe (data not shown), in agreement with the Randles-Sevčik equation (Equation (7)).
The second logical step consisted of the specific detection and quantification of AFB1, by the functionalized working electrode. As mentioned in the introduction, three different functionalization protocols were compared; all were based on the self-assembly technique (chosen for its simplicity) and included a biotinylated antibody recognizing AFB1. The most relevant in terms of reproducibility and efficiency consisted of the direct adsorption of the streptavidin-biotin couple on the gold surface previously cleaned, followed by the immobilization of the biotinylated antibody as show schematically in the Figure 9a [51]. Changes in CV diagrams at each step confirm the functionalization (Figure 9b): Adding SA at the surface of the electrode decreases the current's peak intensity by forming a layer limiting the access of the electrolyte to the surface. The following functionalization by bi-

Application of the Homemade Sensor to the Detection of the AFB1 Fungal Toxin
We first applied to the homemade sensor different concentrations of the Fe 3+ /Fe 2+ redox couple in the absence of functionalization on the working electrode. Three concentrations ranging from 1 mM to 10 mM were studied at a scan rate of 100 mV/s: the amplitude of the CV diagrams decreases proportionally to the concentration of the probe (data not shown), in agreement with the Randles-Sevčik equation (Equation (7)).
The second logical step consisted of the specific detection and quantification of AFB1, by the functionalized working electrode. As mentioned in the introduction, three different functionalization protocols were compared; all were based on the self-assembly technique (chosen for its simplicity) and included a biotinylated antibody recognizing AFB1. The most relevant in terms of reproducibility and efficiency consisted of the direct adsorption of the streptavidin-biotin couple on the gold surface previously cleaned, followed by the immobilization of the biotinylated antibody as show schematically in the Figure 9a [51]. Changes in CV diagrams at each step confirm the functionalization (Figure 9b): Adding SA at the surface of the electrode decreases the current's peak intensity by forming a layer limiting the access of the electrolyte to the surface. The following functionalization by biotin and antibody slightly modify the molecular architecture in terms of organization/orientation leading here to a better access of the electrolyte. A last step consists of the adsorption of BSA to block unbound sites, limiting once more the access to the surface.
Such an architecture was able to detect and to quantify the presence of AFB1 injected through a hydrostatic system. The results shown in Figure 9c,d were obtained with an incubation time of 30 min. The successive increase of AFB1 concentration induces an increase in the oxidation current and the absolute value of the reduction one. Such variations are likely due to the direct toxin-antibody interaction, changing the organization of the molecular architecture at the electrode's surface. Figure 9d also shows a linear relationship between the oxidation current and the AFB1 concentration in a range between 50 fg/mL and 10 ng/mL in chloroform. The limit of detection and the sensitivity are estimated to 50 fg/mL and 18 µA/(ng·mL −1 ), respectively [51].
Chemosensors 2022, 10, x FOR PEER REVIEW 13 of 18 otin and antibody slightly modify the molecular architecture in terms of organization/orientation leading here to a better access of the electrolyte. A last step consists of the adsorption of BSA to block unbound sites, limiting once more the access to the surface. Such an architecture was able to detect and to quantify the presence of AFB1 injected through a hydrostatic system. The results shown in Figure 9c,d were obtained with an incubation time of 30 min. The successive increase of AFB1 concentration induces an increase in the oxidation current and the absolute value of the reduction one. Such variations are likely due to the direct toxin-antibody interaction, changing the organization of the molecular architecture at the electrode's surface. Figure 9d also shows a linear relationship between the oxidation current and the AFB1 concentration in a range between 50 fg/mL and 10 ng/mL in chloroform. The limit of detection and the sensitivity are estimated to 50 fg/mL and 18 μA/(ng•mL −1 ), respectively [51].
Finally, AFB1 was detected in more realistic samples: rice milk samples were centrifuged (5000 rpm, 10 min, 22 °C). AFB1 was then injected in the liquid part resulting from the centrifugation and diluted using the liquid phase obtained from the centrifugation of an untreated milk sample. The electrochemical detection of AFB1 was possible in a range of concentration between 0.5 pg/mL and 2.5 pg/mL, with the sensor behaving linearly in this range.

Discussion
The aim of this work was to elaborate on a complete electrochemical immunosensor, including the electronic platform, for the detection of the fungal B1 aflatoxin (AFB1). The first part of this work was devoted to the design of a system made of three SPEs. The choice of their geometry and their composition was based on required electric properties and manufacturing ease. Since the pastes contain molecules, which could interfere with Finally, AFB1 was detected in more realistic samples: rice milk samples were centrifuged (5000 rpm, 10 min, 22 • C). AFB1 was then injected in the liquid part resulting from the centrifugation and diluted using the liquid phase obtained from the centrifugation of an untreated milk sample. The electrochemical detection of AFB1 was possible in a range of concentration between 0.5 pg/mL and 2.5 pg/mL, with the sensor behaving linearly in this range.

Discussion
The aim of this work was to elaborate on a complete electrochemical immunosensor, including the electronic platform, for the detection of the fungal B1 aflatoxin (AFB1). The first part of this work was devoted to the design of a system made of three SPEs. The choice of their geometry and their composition was based on required electric properties and manufacturing ease. Since the pastes contain molecules, which could interfere with the detection, the electrodes were heated. The amount of undesirable organic molecules clearly decreased significantly for each paste as shown by Raman spectroscopy. AFM was also used to reveal the morphology of the electrode surfaces, which were all rough, with roughness dimensions depending on the nature of the electrode. In the case of the Ag/AgCl electrode, the low roughness suggests an overall homogenous surface. This point is important since it has been shown that the less porous the surface, the greater and more stable the charge transfer [66,67].
The next step of the work consisted of the design of an autonomous and cheap electronic system for the control of the sensor and electrochemical measurements. This system was first validated by the application of a potential in a range adapted to the redox couple used later for the electrochemical detection. Measurements showed a linear and reversible current behavior, in agreement with the Ohm's law.
Our homemade system was also compared to a commercial one, showing current variations significantly higher likely due to a higher A R−S /A geom ratio. This is clearly an advantage for further detections. Moreover, a comparison of CV diagrams suggested that the diffusion layer is thinner in our homemade sensor, which consequently reaches equilibrium faster than the commercial one. Finally, the homemade sensor was able to reach almost the perfect electrochemical reversibility at a scan rate of 0.02 V/s, contrary to the commercial system. Unfortunately, this scan rate did not ultimately make it possible to rapidly obtain a stable response from the sensor. In conclusion, our homemade sensor is competitive with the commercial one being characterized by an amplified response, which may lead to an ultrasensitive response during toxin detection.
The last step was the detection of the AFB1 toxin. The gold electrode was functionalized by a molecular architecture previously described and including an antibody recognizing specifically the toxin. The limit of detection and the sensitivity were estimated to 50 fg/mL and 18 µA/ng.mL −1 , respectively, ranking our homemade sensor among the best sensors already developed for the detection of this toxin [51,[68][69][70][71]. However, the real efficiency test consisted of detecting AFB1 in a complex medium. AFB1 was injected at different concentrations in rice milk and detected in the liquid phase obtained after centrifugation of these solutions and dilutions. The test was clearly positive: the electrochemical detection of AFB1 was possible at low concentration of the toxin, the sensor behaving linearly in a range of concentration between 0.5 pg/mL and 2.5 pg/mL.

Conclusions
In this work, we describe the complete elaboration of an electrochemical immunosensor for the detection of a fungal toxin. The first step consisted in manufacturing appropriate electrodes. We developed and characterized a system made of SPEs connected, in a second step, to a homemade electronic system. The performance of the whole platform was then analyzed and compared to a commercial one. Our approach highlighted the importance of each step of the process that can affect the reliability of the final sensor. Finally, after a simple but efficient functionalization of the working electrode, our setup allowed us to detect the toxin in a complex medium, showing the pertinence of the whole process and the possibility to apply this sensor in an agrifood context. Such a system has different advantages: the simple method used to functionalize the working electrode can be simply adapted for the detection of other toxins or biological molecules. The sensor is also rather cheap taking into account the possibility of printing a high number of electrodes at the same time and the low number of molecules required for the functionalization. In an ecological approach, an improvement of the sensor could now be to reuse the active surface, either by gently removing AF from the functionalized surface or removing the entire molecular architecture without destroying electrodes.